BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:χ² and Beyond: Robust Statistics for Hadronisation Tuning
DTSTART:20261119T130000Z
DTEND:20261119T143000Z
DTSTAMP:20261031T162000Z
UID:indico-event-1671@indico.ifj.edu.pl
DESCRIPTION:Speakers: James Whitehead (Jagiellonian University)\n\nMonte C
 arlo event generators used in LHC phenomenology combine perturbative model
 ling of hard processes with semi-empirical models of non-perturbative phys
 ics\, including hadronisation and the underlying event. These models intro
 duce parameters that must be calibrated\, or 'tuned'\, to data. Standard t
 uning approaches minimise a weighted χ² objective\, so large residuals c
 an exert disproportionate influence. Yet such discrepancies are expected t
 o arise from imperfect phenomenological models\, and may reflect structura
 l limitations that cannot be resolved by adjusting the parameters. Consequ
 ently\, a tune can be driven by precisely the measurements which the model
  is least able to describe.\nThe concept of ‘robustness’\, developed i
 n the statistical framework of Huber and others\, offers a principled way 
 to limit this sensitivity. Robust statistics concerns the stability of inf
 erence under departures from assumed statistical distributions\, including
  contamination by observations drawn from a different distribution. We com
 pare χ² with robust alternatives\, both through loss-function minimisati
 on and Bayesian inference with MCMC sampling. We apply these approaches to
  the final stage of a recent Herwig 7 tune using the Pythia Lund string mo
 del\, and assess how these statistical choices affect the inferred paramet
 ers\, their uncertainties\, and sensitivity to tuning weights.\n\nhttps://
 indico.ifj.edu.pl/event/1671/
URL:https://indico.ifj.edu.pl/event/1671/
END:VEVENT
END:VCALENDAR